A year ago we had 200 users and a Notion doc where we tracked feature requests. Today we have 4,800 designers across 60 countries, a Discord with 14 active channels, and a weekly office hour that regularly runs forty minutes over time because someone always has one more question about token naming conventions. Here is what we've learned.
How we listen
We have three formal listening channels: weekly office hours (open video call, no agenda, bring your tokens), monthly community surveys (eight questions, takes four minutes, 34% completion rate which is absurdly high for a survey), and a Discord thread called #hot-takes where people say things they wouldn't say in a product review.
The #hot-takes channel has been more useful than the surveys. Formal research produces formal answers. People in #hot-takes say things like "the export modal feels like it was designed by someone who has never exported anything" — which is direct, accurate, and led to a complete redesign of the export experience in Q1.
The most useful feedback we've ever received came from a designer who said our token naming suggestions were "confidently wrong." She was right.— Community retrospective, March 2026
The onboarding wall
The single most common point of confusion, across every cohort we've surveyed, is the distinction between primitive tokens, semantic tokens, and component tokens. Designers who come from a Figma-only background understand variables, but the three-layer abstraction of a proper token system is not obvious.
We've tried explaining this five different ways: diagrams, analogies (the operating room one was mine), interactive tutorials, video walkthroughs. The most effective by far was adding a small inline explainer that appears the first time you create a semantic token, with a one-sentence answer to "why not just use the primitive directly?"
- 62% of new users reached the "first token created" milestone in the first session — up from 38% before the inline explainer.
- Support requests about the three-layer model dropped 44% month-over-month.
- The #help channel in Discord saw a 30% reduction in "I'm confused about semantics" threads.
The z-index debate
Seven Discord threads. Seven. All about z-index tokens. The core dispute: should z-index values be named semantically (z.modal, z.tooltip, z.overlay) or numerically (z.100, z.200, z.300)?
The semantic camp argues that names carry intent — if you name a token z.modal, it's obvious what it's for. The numeric camp argues that semantics break down the moment you have two modal-level elements that need different z-indices, and you end up with z.modal-but-for-the-one-behind-the-other-modal.
We landed on a hybrid: a numeric scale (100 through 900 in steps of 100) with optional semantic aliases layered on top. Teams that want semantic names can add them. Teams that want pure numerics can skip the aliases. Both export correctly to CSS. The Discord threads have mostly stopped.
What people actually export
We track export format usage (aggregated, anonymized). CSS Variables is first by a wide margin — 71% of all exports. SCSS is second at 18%. LESS is 6%. Tailwind and Style Dictionary together account for 5%, which surprised us given how much those communities talk about tokens.
More interesting: 34% of all exports are of color tokens only. The full token set (colors + type + spacing + radius + shadows) accounts for only 22% of exports. Most teams are still treating UISqueezy as a color management tool and manually handling the rest. This tells us we have more work to do in explaining the value of the full system.
What comes next
The clearest signal from this year of community calls: people want collaborative review. Right now, a token change in UISqueezy Studio is made by one person and either approved by the pipeline or shipped. There's no human review step built into the workflow.
We're building that now: token change requests, a review UI, comment threads on individual token diffs. It's a smaller version of what GitHub does for code, applied to design decisions. The design-system teams who will benefit most are the ones where the brand team and the engineering team have different opinions about what "brand blue" should be. Which is all of them.